The AERosol RObotic NETwork (AERONET) uses Cimel sunphotometers (AERONET, 2024) and has a centralized data processing and distribution system providing instrument calibrations and standardized data acquisition to retrieve aerosol optical, microphysical, and radiative properties through ground-based passive remote sensing. AERONET Cimel sun photometers are calibrated at the high-altitude station in Izaña by the Langley plot method (Holben et al., 1998). The direct-sun algorithm of AERONET, using the version 3 processing algorithm (Giles et al., 2019) employed in this work, includes level 2.0 AOD measurements at 340, 380, 440, 500, 675, 870 and 1020 nm and AE retrievals at 440–870, 380–500, 440–675, 500–870 and 340–440 nm. Level 2.0 AERONET data have been utilized for several stations for tracing the plume as presented in Sect. 3.2. The AERONET field instruments are inter-calibrated at Mauna Loa and Izana Langley calibrated reference instruments resulting in AOD uncertainty at optical airmass 1 of ∼0.01 in the visible and near infrared increasing to ∼0.02 in the UV wavelengths (Eck et al., 1999). AERONET inversion products from level 2.0 and level 1.5 (level 1.5 is used only when level 2.0 data is not available and an additional criterion is used in this case to filter the data as described below) have also been utilised in this work including single scattering albedo (SSA), absorption aerosol optical depth (AAOD), absorption Ångström exponent (AAE), volume size distribution (VSD) and refractive index (RI). The AERONET inversion products are based on the sky scan radiance measurements at 340, 440, 675 and 870 nm. The nominal uncertainty in AERONET SSA for AOD at 440 nm above 0.40 is ∼0.03 and the SSA uncertainty decreases with increase in AOD (Dubovik et al., 2000; Sinyuk et al., 2020). The estimated uncertainty for real refractive index is well below 0.04 for AOD greater than 0.5 for biomass burning cases (Dubovik et al., 2000; Sinyuk et al., 2020). The uncertainty in volume median radius is estimated to be 0.1±0.5 % and 5.4±4 % for fine and coarse mode, respectively (it is an indication of good confidence in size distribution retrievals) (Sinyuk et al., 2020). The absorption AOD values as reported in the AERONET almucantar inversion dataset are obtained using the relationship AAOD =(1-SSA)×AOD. A detailed uncertainty estimation of AERONET inversion product for version 3 can be found at Sinyuk et al. (2020). For the analysis in this manuscript, we have used the AERONET inversion products Level 2.0 and Level 1.5 (with sky error below 5 % and solar zenith angle above 45° and the coincident AOD values at 440 nm above 0.40).

Precision filter radiometers (Wehrli, 2005), part of the Global Atmospheric Watch-Precision Filter Radiometer (GAW-PFR) network, perform aerosol optical depth measurements at four wavelengths namely 368, 412, 500, and 862 nm (Wehrli, 2000). The primary calibration is performed using the reference triads of PFRs which are calibrated regularly at the high-altitude stations of Mauna Loa in Hawaii, USA and Izaña in Tenerife, Spain using the Langley calibration method (Nyeki et al., 2015). Instruments are calibrated yearly or in 2 years on the basis of instrumental and logistic aspects and the calibration uncertainty is assured to be within ±1 % (Kazadzis et al., 2018b). Quality-assured AOD data are obtained after pre- and post-deployment calibrations. Post-deployment calibrations are done either by the World Optical depth Research and Calibration Center (WORCC) issued calibration certificates or by Langley site calibrations (Kazadzis et al., 2018b; Toledano et al., 2018). For this analysis, we have used the PFR AOD and AE data for Davos (DAV) (46.80° N, 9.80° E, 1580 m above sea level (a.s.l.)) and Jungfraujoch (JFJ) (46.32° N, 7.59° E, 3580 m a.s.l.). AERONET/CIMEL and PFR instruments have been repeatedly intercompared in various studies in order to ensure homogeneous retrievals (e.g., Kazadzis et al., 2018a; Karanikolas et al., 2022).

In addition to the sun photometers and filter radiometers as mentioned above, spectral measurements have also been used. The transportable reference spectroradiometer Quality Assurance of Spectral Ultraviolet Measurements (QASUME) (Gröbner et al., 2005) consists of a scanning double monochromator with a full width at half maximum (FWHM) of 0.86 nm and measures over the spectral wavelengths from 280 to 550 nm. The whole system resides in a temperature-controlled enclosure to allow outdoor operation under varying ambient conditions. The solar radiation is collected with a temperature stabilised diffuser connected via an optical fiber to the entrance slit of the monochromator. A portable lamp monitoring system allows for the calibration of the whole system while being deployed in the field (Gröbner et al., 2017; Hülsen et al., 2016). A collimator tube with a full opening angle of 2.5° is mounted on an optical tracker to which the diffuser head can be fitted, allowing the measurement of direct solar spectral irradiance. The measurements of direct solar spectral irradiance have a standard relative uncertainty of less than 1 %, resulting in a standard uncertainty of 0.014 at 310 nm to 0.007 at longer wavelengths in AOD at an airmass of 1.5 (Gröbner et al., 2023).

The BiTec Sensor (BTS) instruments, manufactured by Gigahertz Optik GmbH, are a system composed of two array-spectroradiometers. The spectral range from about 320 to 1000 nm is covered with a 2048-pixel Si BTS2048-VL-TEC-WP with a nominal spectral resolution (FWHM) of 2.5 nm whose characterization was described in Zuber et al. (2018a, b), while the spectral range from 1000 to 2150 nm is measured with a BTS2048-IR-WP with a nominal spectral resolution of 8 nm (FWHM) with 512 pixel and an extended InGaAs detector. Each spectroradiometer has a collimator to measure direct solar spectral irradiance. The instruments used in this analysis were calibrated at the Physikalisch-Meteorologisches Observatorium Davos and World Radiation Center using the same portable lamp system as described above for QASUME, resulting in similar uncertainties in spectral irradiance and AOD as for QASUME (Gröbner et al., 2023). An estimated expanded measurement uncertainty from 300 to 330 nm is within 3.5 %, 330 to 450 nm is 1.9 % and for the remaining spectral range it is 1.8 % (Gröbner et al., 2023).

Smoke plume optical, microphysical and chemical particle properties were analyzed using the comprehensive in situ instrumentation at the High-Altitude Research Station JFJ. The aerosol measurements of the JFJ site are part of the GAW program, the pan-European Aerosol, Clouds and Trace Gases Research Infrastructure (ACTRIS) and the Swiss National Air Pollution Monitoring Network (NABEL). This instrumentation is briefly described here.

Aerosol light scattering was measured with an integrating Nephelometer (TSI 3563) which provides the total scattering and backscattering coefficients at 450, 550 and 700 nm wavelengths. The instrument output was corrected for angular truncation using the network recommendations (ACTRIS-CAIS-ECAC, 2024), which rely on the scheme from Anderson and Ogren (1998). The measurement reproducibility for submicron data is within ±1 % while that for super micron data lies between 5 %–10 % (Anderson and Ogren, 1998). Light absorption coefficients at seven wavelengths (370, 470, 520, 590, 660, 880 and 950 nm) were determined by a dual spot Aethalometer (MAGEE scientific AE33). The network recommended instrument data processing algorithms based on the work of Drinovec et al. (2015) were applied to correct for non-idealities in filter attenuation-based instrument. The sensitivity of the absorption coefficient determination from AE33 to scattering is below 1 %–1.5 %.

In addition to optical properties, the particle size distributions ranging from 0.01 to 0.8 and 0.5 to 20 µm, were measured by a mobility particle size spectrometer (Wiedensohler et al., 2012), an aerodynamic particle sizer (APS, TSI 3321), and a white light optical particle sizer (Palas Fidas^® 100), respectively. Merged particle size distributions covering the diameter range from 0.01 to 10 µm were generated by joining the MPSS size distributions below 0.6 µm with the averages of the APS and Fidas size distributions above 0.6 µm (after the APS measured diameters were shifted from aerodynamic to mobility diameters assuming an effective density of 1.6 g cm⁻³). The limits formed by the APS and Fidas size distributions around their averages were retained as an indication of the uncertainty in the merged size distributions above 0.6 µm diameter.

A condensation particle counter (TSI 3772), with a lower cut-off size of 10 nm provided the total particle number concentrations. The chemical composition of the non-refractory particulate matter, with aerodynamic diameters smaller than 1 µm was further analyzed in situ and in real time with a Time-of-Flight Aerosol Chemical Speciation Monitor (TOF-ACSM, Aerodyne Inc.) that provides the linear detection of sulphate, nitrate, ammonium, chloride and organic aerosol species through a two-step thermal vaporization (approximately 600 °C) and electron impact ionization process. More details on this instrument can be found in Fröhlich et al. (2013, 2015).

The RAman Lidar for Meteorological Observations (RALMO) was designed by MeteoSwiss and the École Polytechnique Fédérale de Lausanne (EPFL) and is operated at the MeteoSwiss station of Payerne (PAY; 46.80° N, 6.93° E, 492 m a.s.l.), Switzerland, since the year 2007. It provides, continuous operational measurements of humidity since 2008 and temperature and aerosol backscatter since 2010. RALMO is fully automated and operates continuously except in the presence of precipitation or a cloud ceiling below 1000 m above ground level. Data are processed in near-real-time and made available to the MeteoSwiss database and to the international database of Network for the Detection of Atmospheric Composition Change (NDACC), Global Climate Observing System (GCOS) Reference Upper-Air Network (GRUAN) and European Aerosol Research Lidar Network (EARLINET). RALMO uses a narrow field-of-view, narrowband configuration, a UV laser at 355 nm, and four telescopes with 30 cm diameter, fiber-coupled to two grating polychromators for the retrieval of water vapour, temperature and aerosol backscatter. The optical design of RALMO allows the retrieval of the water vapor and the temperature within the troposphere up to the tropopause, while the aerosol backscatter can be retrieved up to the middle stratosphere allowing the detection of elevated stratospheric aerosol plumes. A more detailed description of the aerosol, temperature and humidity transceiver systems is provided in Martucci et al. (2021), and Dinoev et al. (2013) and Brocard et al. (2013), respectively. The estimated error in aerosol backscatter by RALMO at 355 nm for medium-high-aerosol-content data is -0.018±0.237 and +0.001±0.141 Mm⁻¹ sr⁻¹ (+6±38 %) for altitudes 0.8–3 and 0.8–6 km a.s.l., respectively (Brunamonti et al., 2021).

CHM15k ceilometer data has been used for vertical profile observations at Davos which is a one-wavelength backscatter lidar at 1064 nm. The vertical profile of backscatter coefficient dataset is obtained from V-Profiles (https://vprofiles.met.no/about/, last access: 25 May 2025) including the attenuated backscatter signal at 1064 nm from 0 to 6000 m above ground level. The relative error in backscatter coefficient at 1064 nm is 10 % (Wiegner and Geiß, 2012).

The long-range transported smoke from several wildfires burning in Canada in September 2023 reached the Swiss Alps in late September (reaching JFJ on 30 September 2023) and beginning of October (reaching DAV and Payerne (PAY) on 1 October 2023) from the northern part of the European mainland after crossing the North Atlantic Ocean (Fig. 1a).

This was further confirmed by the sky-camera images (Fig. 1b) as well as the Ceilometer measurements (Fig. 1c) on 1 October 2023 at DAV, which detected the presence of plumes above 2 km height above ground level, later settling below 2 km during the day of 1 October. The air mass back trajectories at DAV (Fig. 1a) pointed towards the central European aerosol sources originating from the North-East of Canada, where several wildfires were active during the September of 2023 (Byrne et al., 2024; Jain et al., 2024; Chen et al., 2025). The transported smoke plume was also detected by RALMO in PAY on 1 October 2023 as presented in Fig. 1d with the presence of a thick aerosol layer slowly lowering in altitude from 2 to 1 km, the aerosol layer is characterized by a total backscattering ratio value exceeding 4 from 06:00 to 18:00 UTC.

The event was characterized by increased aerosol load from ground-based aerosol measurements on 30 September 2023 at JFJ and 1 October 2023, at DAV. Apart from the elevated AOD values, a rare feature of this long-range smoke plume transport event was the spectral AOD dependence as observed from ground-based measurements at DAV (Fig. 2). The rare observation is that the spectral AOD first increased from 310 nm to approximately 500 nm and then decreased at longer wavelengths i.e., a distinct spectral curvature in spectral AOD variation with maxima around 500 nm could be observed. Figure 2a shows the spectral AOD comparison as captured with multiple filter- and spectro- radiometers on 1 October at 10:00 UTC at DAV that further confirms the relevance of the occurrence and spectral curvature in AOD variation. Similar spectral AOD variation was also observed for remote GAWPFR station of JFJ with the PFR instrument as presented in Fig. 2b that also had spectral curvature in AOD during some time stamps on 30 September 2023. The temporal evolution of the spectral curvature in AOD as presented in Fig. 2c, d shows that at JFJ, the peak at 500 nm was prevalent during morning time of 30 September 2023 while at DAV, the peak at 500 nm was prevalent through the entire day (1 October 2023). In addition, the spectral AOD analysis for PAY, exhibiting similar spectral behaviour, is provided in Sect. 3.2.1 with the available AERONET measurements.

This rare feature in spectral AOD variation can be considered as an extreme case scenario of an extreme event that was observed for the first time in the measurement history of DAV. The AOD variation at DAV from 2004 to 2022 with nearly 2 decades of AOD values being below 0.11 and the cases of AOD being above 0.5 is only 1 % of the total measurements (Fig. C1a–d). Figure C1e shows the peak daily mean spectral AOD values at DAV from GAWPFR measurements during the same period. The daily mean spectral AOD values are calculated ensuring that there is valid measurement at all wavelengths and considering AOD values up to 3 precision points. The highlighted spectral AOD variation and peak in AOD at 500 nm, was observed only on one day – 1 October 2023 in nearly two decades of continuous measurements at DAV.

To further investigate this special spectral AOD curvature, we studied the detailed microphysical and chemical properties of the particles and gases measured in situ in the smoke plume during its passage over JFJ on 30 September and 1 October 2023. These measurements are summarized in Fig. 3 and Table B1. The plume was characterized by elevated concentrations of CO and Benzene in the gas phase (approximately 100–300 ppb and 20–55 ppt, respectively; Fig. 3a; note that benzene concentrations were not measured during the plume peak due to the instrument calibration schedules). The mass concentration of particles with diameters less than 1 µm (PM1) peaked at approximately 30 µgm-3 during the beginning of the plume period and averaged around 1.3 µgm-3 during the remainder of the plume period. These concentrations are drastically greater than the corresponding concentrations before and after passage of the plume (approximately 0.1–0.6 µgm-3). The particles were mostly composed of organic compounds (approximately 0.7 to 12 µgm-3), with only minor contributions from black carbon (approximately 0.04 to 1.4 µgm-3) and the inorganic species sulphate (SO42-; approximately 0.08 to 0.2 µgm-3) and nitrate (NO3-; approximately 0.02 to 0.6 µgm-3).

The smoke particles scattered and absorbed considerable amounts of light. Aerosol scattering and absorption coefficients at 550 nm were up to 366 and 19 Mm⁻¹, respectively, resulting in single scattering albedo (SSA) values at 550 nm of around 0.95, which are similar to the corresponding SSA values before and after the plume period. The smoke particles had rare optical exponents: e.g., scattering Ångström exponents (SAE) from 450–700 nm were approximately 0, absorption Ångström exponents (AAE) from 370–880 nm were approximately 2, and SSA Ångström exponents were approximately -0.10 during the plume period (Fig. 3d). Usually, aged, long range transported wildfire plumes measured at JFJ have AAE values below 1.5 because of bleaching and evaporation of the brown carbon content (Forrister et al., 2015). Such values are closer to the spectral properties of BC-rich fossil fuel combustion plumes that have AAE values approximately 1.0. However, in this special case on 30 September and 1 October 2023, UV absorption (370 nm) was much higher (approximately 5 times) than visible absorption indicating the presence of brown carbon and/or tar balls, which have a strong spectral dependence in the imaginary refractive index (Corbin et al., 2019). The negative SSA Ångström exponents observed for these smoke particles are also remarkably rare. Normally such negative values are only observed at JFJ during Saharan dust events (SDEs; Collaud Coen et al., 2004). During this special event, the rare optical properties of the smoke particles led to a false trigger of the JFJ SDE alert.

In contrast to the highly elevated aerosol mass concentrations in the plume, the number concentrations of particles in the plume were relatively similar to those before and after plume passage (in all cases within a range of 250–750 particles cm⁻³). This observation is explained by the size distributions of the smoke particles, shown as time series in the image plots in Fig. 3e and f, and for selected time periods in Fig. 4. The number size distributions of particles below approximately 500 nm diameter before, during, and after the plume contained multiple modes and were similar in magnitude. However, during the plume, an additional size mode centred at around 600 nm appeared in the particle number size distribution (Fig. 4a). This mode is even more apparent when plotting the distributions as a function of aerosol volume or VSD (Fig. 4b), demonstrating that these particles led to the substantial increases in the total aerosol volume (and mass) concentrations. The presence and location of this large accumulation mode, as well as the relatively lower coarse mode particle concentrations (diameter >1000 nm), are confirmed by two independent sizing measurements (aerodynamic sizing in the APS, optical sizing in the Fidas), demonstrating a high degree of confidence in these results. While a typical particle size distribution for a transported and aged wildfire smoke plume shows a shifted peak within the fine mode, the large shift towards unusually large particle diameters observed here is a very rare event, as also reported by Eck et al. (2023) and Fiebig et al. (2003).

To further investigate the observed spectral dependence of AOD in for this wildfire plume, we plot the spectral dependencies of absorption, scattering and extinction that were measured in situ during the peak of the smoke plume's passage over JFJ. The absorption coefficients have been shifted to the three wavelengths where the scattering coefficients were measured (450, 550, and 700 nm) using the measured AAE values, and the extinction coefficients at these same wavelengths calculated as the sums of scattering plus absorption coefficients. Measurement errors of 10 % are considered for the independently measured scattering and absorption coefficients, and these are propagated to produce the uncertainty estimates for the extinction coefficients. As expected, the in-situ extinction profile is curved with positive gradient below 550 nm wavelength, consistent with the columnar AOD profiles measured for this plume. Figure 5 shows that this curvature is driven by the same curved feature present in the measured scattering profile. In contrast, the measured absorption profile has a negative gradient across the entire spectral range (as also demonstrated by the AAE values substantially greater than unity plotted in Fig. 3). Given the absolute values of the measured scattering and absorption coefficients (which are also reflected by the high SSA values reported in Fig. 3), one can calculate that AAE would need to increase to a value of ∼4, or total absorbing aerosol concentrations by a factor of ∼4, for the absorption profile to have large enough magnitude to affect the shape of the extinction profile (i.e., to remove its curvature). Such increases are well beyond the scattering and absorption coefficient measurement uncertainties. Therefore, the curved spectral AOD and extinction profiles measured in this plume are due to the similarly curved spectral scattering profiles.

To better understand the reasons for the curved spectral scattering profiles, and thus the corresponding curvatures in spectral AOD/extinction, we performed theory calculations constrained with observational data. All particles were assumed to be identical homogeneous spheres (except for their diameter), i.e., all components including BC and BrC were assumed to be present as homogeneous internal mixture with a single, effective refractive index. The polydisperse size distribution was constrained MPSS and APS in situ data, and the complex refractive index was taken from DAV AERONET retrieval results. Figure 6 shows the spectral dependence of absorption coefficient (Fig. 6a), scattering coefficient (Fig. 6b) and extinction coefficient (Fig. 6c) calculated with Mie theory for the polydisperse aerosol as described above. Consistent with the corresponding measurements (Fig. 5), it is clearly seen that the inverted spectral dependence of extinction, i.e., positive gradient below ∼600 nm wavelength, is driven by similar inverted spectral dependence of scattering, and not by the negative gradient in spectral absorption.. It should be noted that if one wished to perform more accurate calculations of the absorption profiles, one would need a model with a more explicit representation of particle morphology and mixing state to account for the presence of internally mixed BC aggregates, which contribute substantially to the measured absorption but only ∼3 %–5 % to the total particle mass in the plume. We refrain from such calculations here since we are primarily interested in the spectral profiles of total particle scattering, extinction and AOD.

In Fig. 6d, e and f we present approximately calculated spectral mass specific absorption, scattering, and extinction cross sections, respectively, for perfectly monodisperse homogeneous spheres with the same complex refractive index as above, to further assess the role of material properties and particle size (Fig. 6g, h, and i present the same calculated curves on a linear y axis scale). Figure 6d and g corroborate that light absorption has a typical spectral behaviour with a negative gradient for all particle sizes. By contrast, the spectral dependence of light scattering (Fig. 6e and h) depends strongly on particle size. The calculated mass scattering cross-section (MSC) has a maximum in the visible for the examples with diameters between 300 and 800 nm, most prominently and most similarly to the observed spectral AOD for the 500 nm example. By contrast, MSC exhibits a strong negative gradient for smaller diameters and a weak positive gradient for larger diameters. So, the observed rare spectral AOD observed in the wildfire plume is for the most part a result of the rather unusual model diameter of the volume size distribution, which exactly falls to the size range, where light scattering shows this feature. Here it is important to stress that the real part of the refractive index was chosen to be independent of wavelength, which is in agreement with the AERONET retrieval results. Light scattering also depends on the imaginary part of the refractive index, which has considerable spectral dependence based on the AERONET retrieval result, in agreement with AAE significantly larger than unity. Therefore, we performed additional Mie calculations with the imaginary refractive index set to a constant value or to zero. The results, which are not shown here, demonstrate, that the spectral dependence of MSC and the mass extinction cross-section is largely insensitive to these changes in the imaginary refractive index. So, we can conclude that the peak of MSC at a wavelength of around 600 to 700 nm for 800 nm particles is a mere particle size effect, i.e., a result of enhanced scattering near the centre of the Mie regime, whereas the drop towards shorter wavelength or longer wavelength is caused by the transitions towards the geometric or Rayleigh regimes, respectively.

We further investigated the track of the plume using ground based AOD measurements by AERONET. The description of the AERONET stations considered in this analysis is presented in Table 1. These stations were selected on the basis of high AOD days within the considered time period, i.e., from 20 September 2023 to 5 October 2023, and on AOD peak occurring at wavelengths other than the shortest AERONET AOD measurement wavelength (i.e., 340 nm). These stations are categorised on the basis of their location, either east or west of the North Atlantic Ocean. During the year 2023, AOD of the days exceeding 3 times the mean AOD of the respective site were selected as high AOD days. For these days, peak AOD occurring at wavelength other than 340 nm (minimum measurement wavelength) was used as the criterion for spectral curvature.

Figure 7a and b present the variation of AOD at 500 nm from 20 September to 5 October 2023, at AERONET stations located west and east of the North Atlantic Ocean, respectively, as described in Table 1. The peak day of the event for each station was selected based on the observed elevated AOD values as described above (LIL, HPB and KZO were stations with less than 3 times mean AOD values but were considered due to observation of the spectral curvature). The corresponding size distributions for each station are presented in Fig. 7a1–b10 (the missing stations here are due to unavailability of VSD data of sufficient quality for reliable interpretation). Several studies (e.g., Zheng et al., 2020; June et al., 2022; Lu et al., 2025) have reported high fraction of particles with large diameter within accumulation size ranges, approximately ranging between 100 to 250 nm for wildfire smoke plume transports. However, the wildfire transport case presented here showed a shifted peak in VSD for submicron particles with diameters close to approximately 800 nm. Moreover, the VSD in coarse mode with almost flat curve in comparison to the sub-micron particle size peak and close to zero was observed at DAV indicating the case of very low or no background aerosols during the plume passage at this station (properties before, during and after the peak day of the plume is presented in Table 2).

Figure 8 shows the spectral variation of AOD on the peak day of the event at all the stations shown in Fig. 7, with an indication of the peak AOD value at one time instance. Unlike the usual spectral AOD variation that monotonically decreases with wavelength, the cases presented here showed a spectral curvature in AOD with peak AOD values occurring at wavelengths other than the lowest AERONET measurement wavelength i.e., 340 nm. The noticeable stations are DAV, HFX and BRT, with a peak in AOD at 500 nm, while other stations showed this special characteristic of spectral AOD peak at 500 nm. DAV and BRT had a very well-defined peak at 500 nm throughout the day.

The mean (standard deviation), maximum and minimum values of the AOD difference (AOD at 500 nm – AOD at 340 nm) is found to be 0.21 (0.07), 0.00, 0.34, respectively for HFX, 0.20 (0.07), 0.06, 0.29, respectively for BRT and, 0.05 (0.01), 0.02, 0.07, respectively for DAV. This indicates that the observed spectral curvature in AOD is significant in comparison to the measurement uncertainty in AOD (∼0.01 in the visible and near infrared increasing to ∼0.02 in the UV wavelengths). An overview of mean size distribution of fine and coarse mode and peak in AOD concave spectral curvature during the peak day of the event at considered stations is provided in Table C1 in Appendix C.

Figure 9 presents the spectral SSA variation during the peak day of the event at the respective AERONET stations and at JFJ from in situ measurements. It is to be noted that the AERONET-inverted SSA values represent effective values averaged over atmospheric columns, and care should therefore be taken when comparing these to the direct, in situ measurements, as well as when thinking about the fundamental factors that affect these AERONET products. The spectral SSA was high (mostly above 0.9) at all wavelengths, increasing in UV-VIS region from 440 to 675 nm, and then remaining either constant or decreasing at longer wavelengths. For dust particles, the SSA typically monotonically increases with wavelengths (Li et al., 2015; Collaud Coen et al., 2004; Yus-Díez et al., 2021; Kaskaoutis et al., 2021) whereas for black carbon dominated absorbing aerosols, the SSA usually monotonically decreases with wavelength (Li et al., 2015; Chauvigné et al., 2019; Kaskaoutis et al., 2021). However, nonmonotonic spectral curvature in SSA is dominant in East Asia, peaking at 675 nm, as reported by Li et al. (2015) and attributed to enhanced absorption by some dust aerosol species in the UV region that become non-absorbing in the visible part of the spectrum and black carbon dominant absorption at longer wavelengths. The uniqueness of the case in the present manuscript is that similar nonmonotonic spectral SSA curvature occurred under conditions of large submicron particles with narrow size distribution (as presented in Figs. 4 and 7) and containing BrC and/or tarballs, as indicated by the high AAE values reported in Fig. 3.

As presented by the authors in Giles et al. (2012), fine mode particles containing hygroscopic aerosol particles such as sulfates have a nearly neutral SSA spectral dependence (Dubovik et al., 2002), while those containing black carbon aerosol particles as the sole absorber exhibit a decreasing SSA with wavelength, and those composed of brown carbon (BrC) or organic carbon (OC) aerosol particles as sole absorber exhibit an increasing SSA with wavelength (strong absorption in ultraviolet and visible region) (Eck et al., 2009). Moreover, a varying concentration of BC with dust, BrC, and/or OC aerosol particles can lead to ambiguous (increasing, decreasing, or constant) SSA dependence on wavelength depending on the spectral absorption properties of the aerosol mixture with the net effect being stronger absorption across the retrieved spectrum (e.g., 440 to 1020 nm) (Dubovik et al., 2002; Giles et al., 2011, 2012).

In the case presented here, the SSA concave spectral curvature for smoke aerosol presents a similar spectral behaviour as observed for absorbing dust aerosols in the UV-Vis region below 675 nm. In the case of dust aerosols, pure coarse mode dust usually shows an increasing SSA with wavelength whereas a mixing of dust with BC or other pollution leads to the SSA concave curvature as presented by Li et al. (2015); however, this concave spectral curvature in SSA (due to absorbing characteristics of coarse mode particles) did not produce spectral curvature in AOD. Another study over the Himalayan region by Tian et al. (2023) also observed a spectral curvature in SSA and scattering coefficient but the absorption coefficient monotonically decreased with wavelength for an aerosol size distribution indicating the presence of both fine and coarse mode particles during the studied biomass burning event. However, in the present manuscript, most of the stations (Fig. 9) showed an SSA concave spectral curvature except DAV and JFJ, which show an SSA monotonically increasing with wavelength (a spectral behaviour similar to pure dust case), and these are associated with the concave spectral curvature in AOD for fine mode aerosols as presented in Figs. 7 and 8.

The case of DAV (from remote sensing measurements) and JFJ (from in situ measurements) as presented in Fig. 9g and i showed a dust-aerosol-like SSA spectral variation, with SSA monotonically increasing with wavelength, for large fine mode smoke aerosols (Fig. 7b). This suggests that in this particular transported plume, the peculiar size distribution with the large fine mode aerosols as well as the presence of BrC leading to high absorption in the UV lead to the peculiar increasing SSA with wavelength.

In this section, we discuss the stations that displayed a peak in AOD at 500 nm, categorizing the transported smoke plume as fresh or slightly aged in case of the stations near the fire source i.e., HFX and BRT, and aged smoke in case of the station located far away from the fire source i.e., DAV.

HFX is a coastal station, BRT is characterized by the presence of hills and mountains in the area and DAV is a high mountain location. The size distribution at DAV was dominated by the fine mode with no discernible coarse mode, while HFX and BRT had a large fine mode and a very small coarse mode (see Fig. 7) during the studied BB event.

Table 2 presents the observed aerosol properties for fresh/slightly aged (at HFX and BRT) and aged smoke (at DAV) plume during peak day of the event. HFX and BRT were characterized by extremely high AOD values during the peak day of the event with almost not much difference (less than 0.01) in negative AOD AE values in wavelength pair of 340–440 and 380–500 nm and a non-monotonic SSA concave spectral curvature with peak at 675 nm. On the contrary, DAV had high AOD (with respect to the regional climatological values) but not of the magnitude as HFX and BRT with approximately 0.02 difference in negative AE for the wavelength pair of 340–440 and 380–500 nm and a monotonically increasing SSA with wavelength (which is a typical characteristic of coarse mode aerosol particles and/or aerosols containing strong UV absorbers such as BrC).

Figure 10 shows the absorption AOD and absorption AE averages during the peak day of the event for these three stations discussed in this section, as well as the real and imaginary parts of the refractive indices. For fresh/slightly aged smoke at HFX and BRT, the absorption AOD at 440 nm was approximately 0.14, with absorption AE (AAE) in the wavelength range 440–870 nm being close to 2.0. In contrast, for aged smoke at DAV, the corresponding AAOD and AAE were observed to be lower (close to 0.05) and higher (close to 3), respectively. There is steeper AAOD variation with wavelength on logarithmic scale that corresponds to higher AAE values. The imaginary part of the refractive index values was almost identical for HFX and BRT decreasing from 0.009 from 440 to 675 nm, and slightly increasing afterwards to approximately 0.005 to 1020 nm. In contrast, at DAV it was higher, decreasing from 0.0018 to 0.004 from 440 to 1020 nm.

The real part of the refractive index was highest at HFX (approximately 1.58 at 440 nm), followed by BRT (approximately 1.57 at 440 nm), and lowest at DAV (approximately 1.55 at 440 nm) for the wavelength range between 440–1020 nm. Hence, it indicates the presence of highly absorbing aerosols near the source for fresh/slightly aged smoke plume. However, the long range transported smoke plume had lower absorption AOD (AAOD) and higher contribution of AAOD to AOD at 440 nm (Table 2). The imaginary part of the refractive index (IRI) decreases with wavelength, which indicates the presence of BrC. This is consistent with the AAE from in situ observation being substantially larger than unity as presented in Sect. 3.1.3 (BC with constant IRI has an AAE of around unity). Also, due to the very high SSA at longer wavelength, indicating a very low BC fraction, the SSA also remains high at shorter wavelength despite BrC contributing to absorption.

Figure 10c, d show the variation of extinction AE with different wavelength pairs at 440–870, 380–500, 440–675, 500–870 and 340–440 nm during the peak day of the event at HFX, BRT and DAV. The extinction AE at all wavelength pairs was observed to be much lower than for typical smoke aerosols. The highlight here is the negative AE in wavelengths pairs involving 340 and 380 nm i.e., wavelengths corresponding to the UV part of the electromagnetic spectrum, and low AE with longer wavelength pairs (also shown in Fig. C2 in Appendix C for all stations). This negative extinction AE is associated with the concave spectral curvature in AOD following the logarithmic spectral variation of AOD using the Ångström power law [Ångström, 1929; Martínez-Lozano et al., 1998]. The change in sign of extinction AE for the wavelength ranges shown in Fig. 10c and d is a result of the maxima in AOD at around 500 nm, which is driven by the large particle diameter of the dominant fine mode aerosol.

In order to estimate the favorable physical parameters associated with the peak at 500 nm causing the observed concave spectral curvature in AOD, we looked into the role of refractive index, particle size, wavelength of incoming of incoming radiation on the extinction efficiency calculated from Mie scattering theory as presented in Fig. 10e from remote sensing measurement (also in Fig. 6 with size distributions constrained by in situ measurements). The refractive index and the particle size are considered as presented in Figs. 10 and 7, respectively for HFX, BRT and DAV. For a refractive index in range 1.53–1.56, it is observed that for particle radius of 0.34 µm and wavelength varying from 340–500 nm, the extinction is found to decrease with increase in size parameter which is inversely relation with wavelength (i.e., extinction efficiency increases with wavelength). This direct proportionality of extinction efficiency to wavelength was found to be pertinent for the particle size above about 0.25 µm for submicron particles with refractive index varying in the range 1.53–1.56.

Another point to note in this smoke plume transport is that the fire plume was visible in the MODIS true colour images (Fig. D1 in Appendix D) even in the presence of clouds which indicates that the injection height of this Canadian wildland fire was quite high. Zhang et al. (2024) confirmed high pyrocumulonimbus (pyroCb) activity (a cumulonimbus clouds formed by wildfires-driven convection) activity during the Canadian wildfire season of 2023; however, only a minor amount of aerosol managed to reach above the troposphere, with most convective events limited to the upper troposphere. Black carbon (BC) is one of the main constituents associated with pyroCb-injected smoke aerosols. A study by Beeler et al. (2024) revealed that a distinguishing feature of BC-containing particles associated with pyroCb clouds can develop thick external coatings of >200 nm thick (made of condensed organic matter) and also other studies showed that BC particles coating thickness can vary between <30 to 400 nm (Katich et al., 2023). Beeler et al. (2024) also reported that pyroCb BC (approximately 18.3) has large average ratio of coating mass to BC mass as compared to the urban (approximately 1.8) and usual wildfire source-based BC (approximately 7.3). A study by Huang et al. (2024) found that atmospheric aging leads to greater heterogeneity in coating-to-BC mass ratios, reshaping particles to be more spherical and compact, and the heterogeneity in size, coating and non-spherical shapes may account for about 20 % and 30 % of the observed reduction in BC absorption capacity, respectively. Another study by Dahlkötter et al. (2014) revealed a quite concentrated transatlantic plume probe over Europe, for which BC was shown to be heavily coated, which indicates photo-chemical production and condensation of secondary organic aerosol along the transport. The evolution of such coated BC particles occurs also in other settings (e.g., urban outflows), and has important implications that can be considered in calculations of the direct effects of aerosols on climate (Tiwari et al., 2023; Liu et al., 2024c; Lee et al., 2022), as well as when considering the relationships between BC emissions (and associated policy actions), BC properties and remotely-sensed optical properties (Liu et al., 2024a, b). While particle size appears to explain the observed spectral curvature in AOD, other factors including potential coating-induced enhancement or non-linear effects on multiple scattering might also be considered for future studies that may or may not be fully separable given the measurement uncertainties.

Similar special spectral AOD variation was observed in another wildfire event in California/Oregon in 2020 as presented by Eck et al. (2023) where the authors described the strong presence of coated black carbon and/or BrC in these transported plumes that stayed over the Pacific Ocean for some time before turning back to the land surface. HFX and BRT stations of the present manuscript seem to be similar to the Eck et al. (2023) case, while DAV is special concerning the monotonically increasing SSA with wavelength despite no visible signatures of coarse mode particles.

Such extreme size distributions have been observed in some of the past volcanic events, which led to the appearance of blue/green sun. A study by Wullenweber et al. (2021) and references therein described that narrow aerosol particle size distributions narrow, centred around a radius of 500 nm, can cause anomalous scattering (i.e., increase in scattering cross sections with wavelength in the visible part of the spectrum). This work was associated with the appearance of blue coloured sun post volcanic eruption (e.g., Krakatao in 1883) or massive forest fires due to Rayleigh scattering as well as minor impacts of water vapour and ozone. Even though the phenomenon of occurrence of blue sun was not observed in the Canadian plume transport analysed in the present manuscript at the extremely special features presented at DAV on 1 October 2023, this was probably the closest extreme condition reached for this phenomenon to occur in the history of measurement of DAV.

Wullenweber et al. (2021) observed that a low refractive index combined with a larger particle radius range led to stronger anomalous scattering and increased the probability of occurrence of blue sun phenomenon. The observed real refractive index in the case of occurrence of blue sun phenomenon as reported by some previous studies by Penndorf (1953) and Ball et al. (2015) was around 1.46 and 1.50 and hence considered to be in the range of 1.3 to 1.5. However, in the case of DAV, the real refractive index was above 1.5 at all wavelengths that might explain the non-occurrence of blue sun phenomenon in this Canadian smoke plume transport. As reported by Wullenweber et al. (2021), unfortunately no aerosol samples are available from the past occurring extreme size distribution events, therefore the true aerosol optical properties associated with such events are unknown. Hence, the aerosol optical properties observed during the Fall 2023 Canadian wildfire smoke plume transport as presented in this manuscript as well as the California/Oregon wildfires smoke plume transport in 2020 as presented by Eck et al. (2023) can serve as a probable sample for future studies exploring this unusually extreme size distribution occurrence and associated anomalies.